[SOLVED] Faster ways to test if a network share is available or not without waiting 15 seconds

Ulysses_ · 09-10-2018, 04:11 PM

Got a network share mounted on directory /mnt/share. If the server where it is located is switched off and a script tries to access a certain file in the share, it fails of course but it fails after 15 seconds. What are some faster ways to test if the share is up and running before trying to access the file, and do so repetitively in a script every second?

Distro: peppermintOS on all systems, based on ubuntu.

ferrari · 09-10-2018, 08:37 PM

Ping the server for a response perhaps? If it is a samba share, maybe a broadcast using 'smbtree -b'? (Maybe not every second though.)

Ulysses_ · 09-11-2018, 06:42 AM

Just found smbtree -b still shows the same output after the server is disconnected, even minutes later. Ping looks more promising as you control the time it waits. Any test for the existence of the share?

pan64 · 09-11-2018, 10:05 AM

probably you can configure timeout settings.

teckk · 09-11-2018, 11:45 AM

Quote:

What are some faster ways to test if the share is up and running before trying to access the file, and do so repetitively in a script every second?

Example, check for machine, if up do something, else not.

Code:

#Check for machine up every 5 seconds
while :; do
    ping -c1 172.16.0.13 &> /dev/null
    if [ $? -eq 0 ]; then 
        echo "Machine is up"
        else echo "Machine is down"
    fi
    sleep 5
done

michaelk · 09-11-2018, 12:29 PM

You never posted what type of share or how it is mounted i.e. via your script or fstab etc. From post #3 we assume it is CIFS. Since you are using peppermint why not nfs or maybe sshfs or maybe use autofs or systemd automount.

Depends on the distribution but you can test to see if the share is available using smblient or maybe avahi-browse. Typically one might assume that if the server is powered on that the samba will automatically start?

Ulysses_ · 09-11-2018, 02:29 PM

Using CIFS only because it's preconfigured in the distro. I mount it with fstab, and if the server is down and "ls /mnt/shared" fails, and then the server is powered on, then "ls /mnt/shared" works. Might the other options report unavailability of the share faster?

Anyway, ping as in teckk's code is probably less wasteful of bandwidth. And it can have -W 2 added to give up waiting for a response after 2 seconds when the server is down. Tried avahi-browse -a and it does not exit, is this normal? Turning off the server, and running avahi-browse -a again reports exactly the same lines.

Ulysses_ · 09-11-2018, 02:40 PM

Is there an easy way for the server to tell other systems that a file in the share has changed so they do not need to poll every second?

ferrari · 09-11-2018, 02:50 PM

Quote:

Tried avahi-browse -a and it does not exit, is this normal? Turning off the server, and running avahi-browse -a again reports exactly the same lines.

The avahi-daemon employs caching. AFAIU, only restarting the daemon will flush the cache.

ferrari · 09-11-2018, 02:55 PM

Quote:

Is there an easy way for the server to tell other systems that a file in the share has changed so they do not need to poll every second?

Not that I'm aware of.

Ulysses_ · 09-12-2018, 12:41 PM

Maybe the process that changes the file could send a UDP packet that a process on the other side is listening for? Is there a command that listens on a port and executes a script when a packet arrives?

EDIT: There is indeed, with short messages instead of UDP packets:

socat -u tcp-l:7777,fork system:./getmsg.sh

where getmsg.sh is the script to execute when a message arrives on port 7777:

#!/bin/sh
read MESSAGE
echo "$MESSAGE"

And now the process (actually a script of mine) that changes the file can notify the other side like this:

echo "hello" | netcat 10.0.0.3 7777

where 10.0.0.3 is the IP of the listening side.